Colorado Repeals, Kansas Litigates: K–12 AI Governance Returns to District Hands

Framing

Last week, the most consequential governance signal for public K-12 leaders came from a federal courthouse. This week, it came from a state capitol. On May 14, 2026, Colorado Governor Jared Polis signed Senate Bill 189, repealing the Colorado Artificial Intelligence Act, the most ambitious state-level AI law in the country, and replacing it with a narrower disclosure-based framework. The replacement statute explicitly carves out FERPA-covered educational institutions, as well as HIPAA-regulated entities, insurers, creditors, and FDA-regulated medical devices. Two structural conclusions follow for public education.

First, the regulatory ground is shifting under districts even before most have built basic governance architecture. Colorado spent two years building a duty-of-care, risk-management, and impact-assessment regime that vendors and counsel had begun to treat as a national reference. That regime is now gone, replaced by a disclosure-based model that imposes a far lighter compliance burden on AI developers and deployers. The bill passed with bipartisan margins (Senate 34-1, House 57-6) and goes into effect January 1, 2027. Other states watching Colorado as a template will now have cover to pull back on prescriptive AI governance frameworks. Districts that delayed AI policy work, hoping that state law would do the heavy lifting, will be governing in a thinner regulatory environment than they expected.

Second, the federal posture continues to push districts to act on AI while limiting the constraints around that action. The Department of Education Supplemental Priority on Advancing AI in Education became effective May 13, 2026. The Justice Department AI Litigation Task Force, established under the December 2025 executive order, became operational on January 10, 2026, with a mandate to challenge state AI laws in federal court. The signal is consistent across federal and state actions in May 2026: districts will be expected to adopt AI, will be eligible for federal grants premised on AI integration, and will be governed primarily by federal civil rights, privacy, and procurement law rather than by state AI statutes. The institutional architecture inside districts is where the legal exposure now sits. That architecture must be built locally because no higher authority will build it for you.

Top Research and Policy Signals

1. Colorado repeals and replaces its landmark AI Act; statute now includes FERPA carve-out

Source type. State legislation (signed by Governor May 14, 2026).

Colorado SB 189 repeals the Colorado AI Act (CAIA) and replaces it with a narrower disclosure-based framework for high-risk automated decision-making systems. The repealed law had imposed a duty of care, mandatory risk-management programs, algorithmic impact assessments, and broad notice obligations on developers and deployers of high-risk AI systems, including in education-adjacent contexts. The replacement statute drops the duty-of-care and impact-assessment requirements. It adds sector-specific accommodations for HIPAA-covered entities, FERPA-covered educational institutions, insurers and creditors, and FDA-regulated medical devices. The bill passed by a bipartisan 34-1 Senate vote and 57-6 House vote and becomes effective January 1, 2027.

Leadership implication. Districts in states that were waiting on Colorado as a template should reset their assumptions. The state-law backstop for AI governance just got thinner. K-12 cabinets should not interpret the FERPA carve-out as relief; FERPA was never designed to regulate algorithmic decision-making, and the carve-out shifts responsibility back to the district level. Boards should ratify a written AI governance position this summer that does not rely on state law for its substance, and procurement language should require vendors to certify compliance with district-specific governance obligations regardless of state statute.

2. Michigan Department of Education releases state AI guidance for districts

Source type. State agency guidance (issued May 12, 2026).

The Michigan Department of Education released formal AI guidance on May 12, 2026, structured around six essential practices: keeping AI purposeful and safe, protecting privacy and integrity, building AI literacy, maintaining human oversight, promoting equity and accessibility, and supporting transparency and continuous improvement. The release includes a starter guide and a comprehensive framework for district policy development. Michigan joins approximately thirty-five states with official AI guidance for K-12 schools, though the substantive depth and enforceability of state guidance documents vary widely.

Leadership implication. A state guidance document is not a district AI policy, and treating it as one creates a false sense of compliance. Michigan superintendents should use the MDE framework as a checklist against their own board-ratified policy, not as a substitute for that policy. In non-guidance states, cabinets shouldn't wait for their agency to act; the gap between agency guidance and district-level operational policy is where civil rights exposure, procurement risk, and instructional incoherence accumulate.

3. CRPE updates Early Adopters database: 79 districts now tracked, up from 40

Source type. Institutional research report, not peer-reviewed.

CRPE updated the Early Adopters database, which tracks 79 U.S. school districts with publicly visible AI strategies for the 2025-26 school year, up from 40 the prior year. Professional development for teachers is the fastest-growing category, present in 86 percent of tracked districts (up from 63 percent), followed by published guidance (78 percent versus 65 percent), teacher tool support (77 percent versus 70 percent), and student tool support (63 percent versus 58 percent). Roughly 37 percent of tracked districts now implement five or more AI initiatives. CRPE characterizes the field as broader but still fragmented, with adoption outpacing systemic coherence.

Leadership implication. Doubling the population of publicly visible AI-adopting districts in one year does not mean those districts have a coherent governance position. Strategic coherence is the variable boards should ask about, not initiative count. Superintendents should be able to answer three questions in writing this summer: who owns AI governance at the cabinet level, what the policy is on AI in student-facing decisions, and what the policy is on AI in employment-related decisions. Without those three answers on paper, the district remains exposed regardless of how many AI initiatives are underway.

4. Federal litigation against AI student-surveillance software continues to advance

Source type. Federal court filings and contemporaneous reporting (not peer-reviewed).

In the ongoing federal case brought by nine current and former students against Lawrence Public Schools (USD 497) in the District of Kansas, U.S. District Judge Kathryn Vratil issued early-motion rulings ordering the district to comply with Kansas Open Records Act requests related to its prior use of Gaggle and its subsequent contract with ManagedMethods. The plaintiffs allege that broad AI surveillance of students' electronic communications and documents violated their Fourth Amendment rights against unreasonable searches and seizures and their First Amendment rights to free expression. The case remains pending on the merits. Education law counsel are now publicly analyzing AI student-monitoring platforms as a constitutional litigation surface, not only a privacy or vendor-management surface.

Leadership implication. Districts running AI-driven monitoring of student communications should treat this case as an early signal of the litigation surface they now occupy. Three positions should be on paper before the next school year: a written justification for the educational purpose of any AI monitoring tool, a documented review process before any flagged content is acted on, and a clear records-management posture aligned with state open-records and FERPA obligations. Switching vendors does not extinguish the constitutional question; what the system does to student communications is what triggers Fourth and First Amendment scrutiny.

5. Cognitive offloading through digital tools: peer-reviewed evidence that self-efficacy mediates learning depth

Source type. Peer-reviewed, published in Frontiers in Psychology, March 12, 2026.

This peer-reviewed study examines the relationship between cognitive offloading to digital tools (including AI) and three downstream outcomes: critical thinking, task persistence, and depth of learning. The authors find that the effect is conditional rather than uniform: students with stronger cognitive self-efficacy use digital tools in ways that support deeper engagement and persistence, while students with weaker self-efficacy show the dependency pattern that has dominated public concern. The study is featured here, rather than newer preprints, because it is one of the few peer-reviewed pieces in the current cycle that directly addresses the critical-thinking concern now central to parent, board, and policy conversations about AI in classrooms.

Leadership implication. District AI literacy programs should not treat AI as a uniform cognitive risk or benefit. The mediating variable is student self-efficacy, which is a teachable construct. Curriculum and PD leaders should specify, in their AI literacy scope and sequence, where students build the metacognitive and self-regulation skills that determine whether AI use supports or undermines learning depth. Vendor pitches promising universal cognitive gains from AI tools should be evaluated against this evidence and discounted accordingly.

Emerging Strategic Themes

Theme 1 — State regulatory retrenchment is now a real variable. The Colorado repeal will not be the last. States that enacted broad AI governance frameworks under political pressure in 2024 and 2025 are facing industry resistance, signals of federal preemption, and pressure from the Justice Department's AI Litigation Task Force. K-12 cabinets that assumed state law would carry substantive governance obligations should recalibrate. Local district policy is now the operational regulatory layer.

Theme 2 — AI monitoring of students is becoming a constitutional question. The Lawrence federal case, combined with parallel scrutiny of AI proctoring and AI behavior analytics across multiple districts, is moving AI student surveillance from a privacy-policy question into a Fourth and First Amendment question. Vendor contracts written before 2026 likely do not contemplate the constitutional defense lift that this category of tools may now require.

Theme 3 — State agency guidance is filling a gap that was never designed for districts to inherit. The Michigan May 12 release is the latest in a wave of state-level AI guidance, but agency documents are designed to inform district policy, not replace it. The substantive policy obligation still sits with boards, superintendents, and cabinets. Treating state guidance as compliance creates the false sense of safety that civil rights plaintiffs will exploit.

Theme 4 — Adoption breadth continues to decouple from strategic coherence. CRPE's doubling of the number of tracked early-adopter districts is a clear signal that AI integration is now mainstream practice, but the same dataset shows that strategic coherence has not scaled with adoption. Boards should distinguish between activity counts and governance maturity. Asking how many AI initiatives are running is a far weaker question than asking who is accountable when one of them fails.

What Was Not Found

Within the May 17 to May 24, 2026 publication window, the following remained absent from the peer-reviewed and high-credibility preprint literature:

No peer-reviewed study published this week measured the accuracy, false-positive rate, or equity effects of AI-driven student monitoring software (such as Gaggle, GoGuardian Beacon, Bark for Schools, or ManagedMethods) on representative U.S. K-12 student populations, despite the constitutional litigation now bearing on this category of tools. The gap is acute: districts are signing renewal contracts, plaintiffs are filing constitutional claims, and the underlying evidence based on monitoring software accuracy has not been independently validated at scale.

No peer-reviewed empirical study published this week measured the comparative effect of state-level AI guidance on actual district policy adoption or student outcomes. With approximately thirty-five states now issuing guidance, the natural-experiment data exists; the rigorous evaluation does not.

No peer-reviewed cost-effectiveness analysis of AI tutoring at scale in U.S. K-12 public school contexts appeared this week. The frequently cited Harvard undergraduate physics RCT, the UK secondary mathematics LearnLM trials, and the Tutor CoPilot mathematics study do not substitute for U.S. K-12 cost-effectiveness evidence.

No peer-reviewed equity audit of state AI guidance variation published this week documented how the patchwork of state-level AI documents differentially serves districts in low-capacity states or in high-poverty contexts. The structural inequity question, that smaller districts get less guidance and less capacity to act on the guidance they receive, remains unaddressed in the peer-reviewed literature.

No peer-reviewed governance maturity instrument published this week could serve as an audit-grade benchmarking tool for superintendents and boards. The instrument-grade work in AI literacy (such as Gao & Wang 2026 for teachers) has no equivalent for board-level governance maturity, despite growing policy demand for it.

The pattern that has defined this brief for months persists. Districts are being asked to make institutional decisions under regulatory and litigation pressure that the peer-reviewed evidence base does not yet support. The governance architecture must be built locally because the evidence will not be available externally within the timelines districts operate on.

Watch This Week

• Ohio district deadline (July 1, 2026) for adopting AI usage policies under S.B. 1734. Districts behind schedule should treat the next forty days as a hard window.
• Continued movement on the K–12 AI Literacy and Readiness Act of 2026 (Rep. Fine) and the bipartisan LIFT AI Act in the House, signaling federal interest in allowing existing education funds to be used for AI literacy.
• Pending merits rulings in the Lawrence Public Schools (USD 497) federal case against AI student-surveillance software in the District of Kansas.
• NSF AI Education Act of 2026 (S. 3957) conference negotiations as federal AI workforce funding moves toward final form.
• Continued litigation under the December 2025 executive order to preempt state AI laws, which now has a live federal force following Colorado SB 189 signing and Georgia and Iowa chatbot bill signings in May.
• Additional state guidance documents expected through June as state agencies respond to the Department of Education's May 13 Supplemental Priority becoming effective.

Sources

Governance and Policy

Research — Peer-Reviewed

Research — Institutional Report (Not Peer-Reviewed)